1. Field of the Invention
The present invention relates to a technology for detecting a specific region such as a human face, and performing different image processing to the detected specific region and other regions.
2. Description of the Related Art
In recent years, realization of high-resolution moving image data has fostered the development of products handling a high-definition (HD) image of, for example, 1920×1080 pixels. Such a trend to realize a larger number of pixels is expected to proceed further, and there is prospect of pixels of 4000×2000 or more in the future.
As a device processing high definition moving image data, a digital video camera supporting an HD image is known. A household digital video camera performs compressing and coding of the moving image data in order to effectively record the HD image with a large amount of information in a recording medium.
As a method for compressing a moving image, the moving picture expert group (MPEG) system (MPEG2 and MPEG4) and H.264 are well known. Some of these compression methods for the moving image reduce code amounts by reducing the image quality in a visually inconspicuous image region, thereby providing high coding efficiency.
However, it is difficult for an apparatus itself to distinguish between a visually conspicuous image region and the visually inconspicuous image region. In most cases, thus, an image is divided into fine pixel blocks to identify the region per pixel block based on characteristics such as presence of an edge and color information.
In a case where a pixel block to be processed has the characteristics of human skin color, the apparatus may deem that there is a high possibility that the pixel block represents a “face”, and accordingly perform control either not to reduce or to increase the code amount generated for the block. In other words, it is a technique to improve an image quality by allocating more code amount to a face region than other regions because deterioration of the image quality is pronounced when noises are generated in a conspicuous region, such as a human face in the image, by compressing and coding the image. Therefore, a technique that can accurately detect whether the pixel block to be processed represents the face region is required.
A method for detecting the face region is discussed in Japanese Patent Application Laid-Open No. 07-087510. Japanese Patent Application Laid-Open No. 07-087510 describes a method for precisely detecting a region along a contour of a characteristics region (face).
However, the method described in the Japanese Patent Application Laid-Open No. 07-087510 separately searches skin color portions and portions adjacent to the skin color, and it must be determined in advance whether pixels around a determination target pixel are skin colored. In other words, determination of the face region has to be performed by two pass procedures which places a heavy processing load on the apparatus. Therefore, a problem that it takes long time to obtain a determination result arises.